⊕ Dataset · slm-data-v2 · v2.0

sciLibRuModal v2

Name: sciLibRuModal v2 · training corpus for SciLibModal
License: on request, with attribution
Keywords: dataset, multimodal, math, lean, latex, image

Training corpus for the SciLibModal model: Mathlib mathematical objects in five modalities (EN, RU, Lean4, LaTeX, image) anchored in the SciLib ontology.

≈ 1 M multimodal records · 5 modalities on request, with attribution

sciLibRuModal v2 is the multimodal training corpus used for SciLibModal. Each record is one mathematical object from Mathlib in five consistent modalities; all modalities are anchored in a single interpretation entity via the SciLib ontology.

Record content: EN/RU statement, Lean signature and body, LaTeX (where generated), formula image (where available). All modalities are parts of one semantic object.

Access: on request via info@scilibai.ru with attribution.

Tags: dataset, multimodal, math, lean, latex, image

← To the catalogue